The evidence that the same total income can lead a household to choose different consumption vectors, depending on who brings in how much of the income, has led to an effort to replace the standard unitary model of the household with the 'collective model', which recognizes that the husband and the wife may have different preferences and depending on the balance of power between them the household may choose differently. One weakness of this new literature is that it fails to recognize that the household's choice could in turn influence the balance of power. Once this two-way relation between choice and power is recognized we, are forced to confront some new questions concerning how to model the household. This paper tries to answer these by defining a 'household equilibrium', examining its game-theoretic properties and drawing out its testable implications. It is shown, for instance, that a household equilibrium can be inefficient and that (for a certain class of parameters) children will be least likely to work in a household where power is evenly balanced. The paper also draws out the implications for female labor supply.